An upgraded Petri net model of systolic architecture for solving differential equations based on Taylor's method

This paper presents an UPN (Upgraded Petri Net) model of a systolic array for solving of differential equations using Taylor's method. The iterative sequence of modelling - simulation - analysis of the model is shown. By evaluating results of the simulation relevant characteristics of the systolic array are observed: synchronism, regularity, time locality and pipeline. Two UPN models are developed. The first one refers to the high level systolic array modeling. This model shows processing elements that calculate or pass their input to the output. Thus, the next processing element use results of the previous one. The second model is more detailed and refers to single processing element. This model is capable to generate solution of differential equation through UPN execution based on Taylor's method. This execution is based on parralel firing of a group of transitions. The suitability of UPN for modelling of the systems at different levels of abstraction is examined and established. Original software for modelling and simulations of Upgraded Petri Nets, PeM (Petri Net Manager), is developed and used for all models described in this paper.